computing the stanley depth
TRANSCRIPT
Journal of Algebra 323 (2010) 2943–2959
Contents lists available at ScienceDirect
Journal of Algebra
www.elsevier.com/locate/jalgebra
Computing the Stanley depth ✩
Dorin Popescu a,∗, Muhammad Imran Qureshi b
a Institute of Mathematics “Simion Stoilow”, University of Bucharest, PO Box 1-764, Bucharest 014700, Romaniab Abdus Salam School of Mathematical Sciences, GC University, Lahore, 68-B New Muslim town Lahore, Pakistan
a r t i c l e i n f o a b s t r a c t
Article history:Received 25 November 2009Available online 16 December 2009Communicated by Steven Dale Cutkosky
Keywords:Monomial idealsStanley decompositionsStanley depth
Let Q and Q ′ be two monomial primary ideals of a polynomialalgebra S over a field. We give an upper bound for the Stanleydepth of S/(Q ∩ Q ′) which is reached if Q , Q ′ are irreducible.Also we show that Stanley’s Conjecture holds for Q 1 ∩ Q 2,S/(Q 1 ∩ Q 2 ∩ Q 3), (Q i)i being some irreducible monomial idealsof S .
© 2009 Elsevier Inc. All rights reserved.
Introduction
Let K be a field and S = K [x1, . . . , xn] be the polynomial ring over K in n variables and M a finitelygenerated multigraded (i.e. Z
n-graded) S-module. Given z ∈ M a homogeneous element in M andZ ⊆ {x1, . . . , xn}, let zK [Z ] ⊂ M be the linear K -subspace of all elements of the form zf , f ∈ K [Z ].This subspace is called Stanley space of dimension |Z |, if zK [Z ] is a free K [Z ]-module. A Stanleydecomposition of M is a presentation of the K -vector space M as a finite direct sum of Stanleyspaces D: M = ⊕r
i=1 zi K [Zi]. Set sdepth D = min{|Zi |: i = 1, . . . , r}. The number
sdepth(M) := max{
sdepth(D): D is a Stanley decomposition of M}
is called the Stanley depth of M . This is a combinatorial invariant which has some common propertieswith the homological invariant depth. Stanley conjectured (see [17]) that sdepth M � depth M , but thisconjecture is still open for a long time in spite of some results obtained mainly for n � 5 (see [1,16,8,
✩ The authors would like to express their gratitude to ASSMS of GC University Lahore for creating a very appropriateatmosphere for research work. This research is partially supported by HEC Pakistan. The first author was mainly supportedby CNCSIS Grant ID-PCE no. 51/2007.
* Corresponding author.E-mail addresses: [email protected] (D. Popescu), [email protected] (M.I. Qureshi).
0021-8693/$ – see front matter © 2009 Elsevier Inc. All rights reserved.doi:10.1016/j.jalgebra.2009.11.025
2944 D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959
2,12,13]). An algorithm to compute the Stanley depth is given in [9] and was used here to find severalexamples. Very important in our computations were the results from [3,6,15].
Let Q , Q ′ be two monomial primary ideals such that dim S/(Q + Q ′) = 0. Then
sdepth S/(
Q ∩ Q ′) � max
{min
{dim
(S/Q ′),⌈dim(S/Q )
2
⌉},min
{dim(S/Q ),
⌈dim(S/Q ′)
2
⌉}},
and the bound is reached when Q , Q ′ are non-zero irreducible monomial ideals (see Proposition 2.2,or more general in Corollary 2.4), � a
2 being the smallest integer � a/2, a ∈ Q.Let Q 1, Q 2, Q 3 be three non-zero irreducible monomial ideals of S . If dim S/(Q 1 + Q 2) = 0 then
sdepth(Q 1 ∩ Q 2) �⌈
dim(S/Q 1)
2
⌉+
⌈dim(S/Q 2)
2
⌉
(see Lemma 4.3, or more general in Theorem 4.5). In this case, our bound is better than the boundgiven by [10] and [11] (see Remark 4.2). Using these results we show that sdepth(Q 1 ∩ Q 2) �depth(Q 1 ∩ Q 2), and
sdepth S/(Q 1 ∩ Q 2 ∩ Q 3) � depth S/(Q 1 ∩ Q 2 ∩ Q 3),
that is Stanley’s Conjecture holds for Q 1 ∩ Q 2 and S/(Q 1 ∩ Q 2 ∩ Q 3) (see Theorems 5.6, 5.9).
1. A lower bound for Stanley’s depth of some cycle modules
We start with few simple lemmas which we include for the completeness of our paper.
Lemma 1.1. Let Q be a monomial primary ideal in S = K [x1, . . . , xn]. Suppose that√
Q = (x1, . . . , xr) where1 � r � n, Then there exists a Stanley decomposition
S/Q =⊕
uK [xr+1, . . . , xn],
where the sum runs on monomials u ∈ K [x1, . . . , xr] \ (Q ∩ K [x1, . . . , xr]).
Proof. Given u, v ∈ K [x1, . . . , xr] \ (Q ∩ K [x1, . . . , xr]) and h, g ∈ K [xr+1, . . . , xn] with uh = vg then weget u = v , g = h. Thus the given sum is direct. Note that there exist just a finite number of monomialsin K [x1, . . . , xr] \ (Q ∩ K [x1, . . . , xr]). Let 0 �= α ∈ (S \ Q ) be a monomial. Then α = u f , where f ∈K [xr+1, . . . , xn] and u ∈ K [x1, . . . , xr]. Since α /∈ Q we have u /∈ Q . Thus S/Q ⊂ ⊕
uK [xr+1, . . . , xn],the other inclusion being trivial. �Lemma 1.2. Let Q be a monomial primary ideal in S = K [x1, . . . , xn]. Then sdepth S/Q = dim S/Q =depth S/Q .
Proof. Let dim S/Q = n − r for some 0 � r � n. We have dim S/Q � sdepth S/Q by [1, Theorem 2.4].Renumbering variables we may suppose that
√Q = (x1, . . . , xr). Using the above lemma we get the
converse inequality. As S/Q is Cohen Macaulay it follows dim S/Q = depth S/Q , which is enough. �Lemma 1.3. Let I , J be two monomial ideals of S = K [x1, . . . , xn]. Then
sdepth(
S/(I ∩ J ))� max
{min
{sdepth(S/I), sdepth
(I/(I ∩ J )
)},
min{
sdepth(S/ J ), sdepth(
J/(I ∩ J ))}}
.
D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959 2945
Proof. Consider the following exact sequence of S-modules:
0 → I/(I ∩ J ) → S/(I ∩ J ) → S/I → 0.
By [14, Lemma 2.2], we have
sdepth(
S/(I ∩ J ))� min
{sdepth(S/I), sdepth
(I/(I ∩ J )
)}. (1)
Similarly, we get
sdepth(
S/(I ∩ J ))� min
{sdepth(S/ J ), sdepth
(J/(I ∩ J )
)}. (2)
The proof ends using (1) and (2). �Proposition 1.4. Let Q , Q ′ be two monomial primary ideals in S = K [x1, . . . , xn] with different associatedprime ideals. Suppose that
√Q = (x1, . . . , xt),
√Q ′ = (xr+1, . . . , xn) for some integers t, r with 0 � r � t � n.
Then
sdepth(
S/(
Q ∩ Q ′))� max
{min
v
{r, sdepth
(Q ′ ∩ K [xt+1, . . . , xn]), sdepth
((Q ′ : v
) ∩ K [xt+1, . . . , xn])}
,
minw
{n − t, sdepth
(Q ∩ K [x1, . . . , xr]
), sdepth
((Q : w) ∩ K [x1, . . . , xr]
)}},
where v, w run in the set of monomials containing only variables from {xr+1, . . . , xt}, w /∈ Q , v /∈ Q ′ .
Proof. If Q , or Q ′ is zero then the inequality holds trivially. If r = 0 then Q ∩ K [x1, . . . , xr] =Q ∩ K = 0, and the inequality is clear. A similar case is t = n. Thus we may suppose 1 � r � t < n.Applying Lemma 1.3 it is enough to show that
sdepth(
Q ′/(
Q ∩ Q ′)) � min{
sdepth(
Q ′ ∩ K [xt+1, . . . xn]), sdepth
((Q ′ : v
) ∩ K [xt+1, . . . xn])}
,
where v is a monomial of K [xr+1, . . . , xn] \ (Q ∩ Q ′). We have a canonical injective map
Q ′/(
Q ∩ Q ′) → S/Q .
By Lemma 1.1 we get
Q ′/(
Q ∩ Q ′) = Q ′ ∩(⊕
uK [xt+1, . . . , xn])
=⊕(
Q ′ ∩ uK [xt+1, . . . , xn]),
where u runs in the monomials of K [x1, . . . , xt] \ Q . Here
Q ′ ∩ uK [xt+1, . . . , xn] = u(
Q ′ ∩ K [xt+1, . . . , xn])
if u ∈ K [x1, . . . , xr]
and
Q ′ ∩ uK [xt+1, . . . , xn] = u((
Q ′ : u) ∩ K [xt+1, . . . , xn]) if u /∈ K [x1, . . . , xr].
2946 D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959
If u ∈ Q ′ then Q ′ : u = S . We have
Q ′/(
Q ∩ Q ′) =(⊕
u(
Q ′ ∩ K [xt+1, . . . , xn])) ⊕(⊕
zK [xt+1, . . . , xn])
⊕(⊕
uv((
Q ′ : v) ∩ K [xt+1, . . . , xn])),
where the sum runs for all monomials u ∈ (K [x1, . . . , xr] \ Q ), z ∈ Q ′ \ Q and v ∈ K [xr+1, . . . , xt],v /∈ Q ′ ∪ Q . Now it is enough to apply [14, Lemma 2.2] to get the above inequality. �Theorem 1.5. Let Q and Q ′ be two irreducible monomial ideals of S. Then
sdepthS S/(
Q ∩ Q ′) � max
{min
{dim
(S/Q ′),⌈dim(S/Q ) + dim(S/(Q + Q ′))
2
⌉},
min
{dim(S/Q ),
⌈dim(S/Q ′) + dim(S/(Q + Q ′))
2
⌉}}.
Proof. If the associated prime ideals of Q , Q ′ are the same then the above inequality says thatsdepthS S/(Q ∩ Q ′) � dim S/Q , which follows from Lemma 1.2. Thus we may suppose that theassociated prime ideals of Q , Q ′ are different. We may suppose that Q is generated in variables{x1, . . . , xt} and Q ′ is generated in variables {xr+1, . . . , xp} for some integers 0 � r � t � p � n. Sincedim(S/Q ) = n − t , dim(S/Q ′) = n − p + r and dim(S/(Q + Q ′)) = n − p we get
n − t −⌊
p − t
2
⌋=
⌈(n − t) + (n − p)
2
⌉=
⌈dim(S/Q ) + dim(S/(Q + Q ′))
2
⌉,
� a2 � being the biggest integer � a/2, a ∈ Q. Similarly, we have
n − p + r −⌊
r
2
⌋=
⌈dim(S/Q ′) + dim(S/(Q + Q ′))
2
⌉.
On the other hand by [6], and [15, Theorem 2.4] sdepth(Q ′ ∩ K [xt+1, . . . , xn]) = n − t − � p−t2 � and
sdepth(Q ∩ K [x1, . . . , xr, xp+1, . . . , xn]) = n − p + r − � r2 �. In fact, the quoted result says in particular
that sdepth of each irreducible ideal L depends only on the number of variables of the ring andthe number of variables generating L (a description of irreducible monomial ideals is given in [18]).Since (Q ′ : v) ∩ K [xt+1, . . . , xn] is still an irreducible ideal generated by the same variables as Q ′ weconclude that
sdepth((
Q ′ : v) ∩ K [xt+1, . . . , xn]) = sdepth
(Q ′ ∩ K [xt+1, . . . , xn]),
v /∈ Q ′ being any monomial. Similarly,
sdepth((Q : w) ∩ K [x1, . . . , xr, xp+1, . . . , xn]) = sdepth
(Q ∩ K [x1, . . . , xr, xp+1, . . . , xn]
).
It follows that our inequality holds if p = n by Proposition 1.4.Set S ′ = K [x1, . . . , xp], q = Q ∩ S ′ , q′ = Q ′ ∩ S ′ . As above (case p = n) we get
sdepthS ′ S ′/(q ∩ q′) � max
{min
{dim
(S ′/q′),⌈dim(S ′/q)
2
⌉},min
{dim
(S ′/q
),
⌈dim(S ′/q′)
2
⌉}}
= max
{min
{r,
⌈p − t
2
⌉},min
{p − t,
⌈r
2
⌉}}.
D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959 2947
Using [9, Lemma 3.6], we have
sdepthS
(S/
(Q ∩ Q ′)) = sdepthS
(S/
(q ∩ q′)S
) = n − p + sdepthS ′(
S ′/(q ∩ q′)).
It follows that
sdepthS
(S/
(Q ∩ Q ′)) � n − p + max
{min
{r,
⌈p − t
2
⌉},min
{p − t,
⌈r
2
⌉}}
= max
{min
{n − p + r,n − p +
⌈p − t
2
⌉},min
{n − t,n − p +
⌈r
2
⌉}}
= max
{min
{n − p + r,n − t −
⌊p − t
2
⌋},min
{n − t,n − p + r −
⌊r
2
⌋}},
which is enough. �2. An upper bound for Stanley’s depth of some cycle modules
Let Q , Q ′ be two monomial primary ideals of S . Suppose that Q is generated in variables{x1, . . . , xt} and Q ′ is generated in variables {xr+1, . . . , xn} for some integers 1 � r � t < n. Thus theprime ideals associated to Q ∩ Q ′ have dimension � 1 and it follows depth(S/(Q ∩ Q ′)) � 1. Thensdepth(S/(Q ∩ Q ′)) � 1 by [5, Corollary 1.6], or [7, Theorem 1.4]. Let D: S/(Q ∩ Q ′) = ⊕s
i=1 ui K [Zi]be a Stanley decomposition of S/(Q ∩ Q ′) with sdepth D = sdepth(S/(Q ∩ Q ′)). Thus |Zi | � 1 forall i. Renumbering (ui, Zi) we may suppose that 1 ∈ u1 K [Z1], so u1 = 1. Note that Zi cannot havemixed variables from {x1, . . . , xr} and {xt+1, . . . , xn} because otherwise ui K [Zi] will be not a freeK [Zi]-module. As | Z1 |� 1 we may have either Z1 ⊂ {x1, . . . , xr} or Z1 ⊂ {xt+1, . . . , xn}.
Lemma 2.1. Suppose Z1 ⊂ {x1, . . . , xr}. Then sdepth(D) � min{r, �n−t2 }.
Proof. Clearly sdepth(D) �| Z1 |� r. Let a ∈ N be such that xai ∈ Q ′ for all t < i � n. Let T =
K [yt+1, . . . , yn] and ϕ : T → S be the K -morphism given by yi → xai . The composition map ψ : T →
S → S/(Q ∩ Q ′) is injective. Note also that we may consider Q ′ ∩ K [xt+1, . . . , xn] ⊂ S/(Q ∩ Q ′) sinceQ ∩ K [xt+1, . . . , xn] = 0. We have
(yt+1, . . . , yn) = ψ−1(Q ′ ∩ K [xt+1, . . . , xn]) =⊕
ψ−1(u j K [Z j] ∩ Q ′ ∩ K [xt+1, . . . , xn]).
If u j K [Z j] ∩ Q ′ ∩ K [xt+1, . . . , xn] �= 0 then u j ∈ K [xt+1, . . . , xn]. Also we have Z j ⊂ {xt+1, . . . , xn},otherwise u j K [Z j] is not free over K [Z j]. Moreover, if ψ−1(u j K [Z j] ∩ Q ′ ∩ K [xt+1, . . . , xn]) �= 0 then
u j = xbt+1t+1 . . . xbn
n , bi ∈ N is such that if xi /∈ Z j , t < i � n, then a | bi , let us say bi = aci for some ci ∈ N.
Denote ci = � bia when xi ∈ Z j . We get
ψ−1(u j K [Z j] ∩ Q ′ ∩ K [xt+1, . . . , xn]) = y
ct+1t+1 . . . ycn
n K [V j],
where V j = {yi: t < i � n, xi ∈ Z j}. Thus ψ−1(u j K [Z j] ∩ Q ′ ∩ K [xt+1, . . . , xn]) is a Stanley spaceof T and so D induces a Stanley decomposition D′ of (yt+1, . . . , yn) such that sdepth(D) �sdepth(D′) � sdepth(yt+1, . . . , yn) because | Z j |=| V j |. Consequently sdepth(D) � �n−t
2 by [3] andso sdepth(D) � min{r, �n−t
2 }.Note also that if t = n, or r = 0 then the same proof works; so sdepth S/(Q ∩ Q ′) = 0, which is
clear because depth S/(Q ∩ Q ′) = 0 (see [5, Corollary 1.6]). �
2948 D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959
Proposition 2.2. Let Q , Q ′ be two non-zero monomial primary ideals of S with different associated primeideals. Suppose that dim(S/(Q + Q ′)) = 0. Then
sdepthS
(S/
(Q ∩ Q ′))
� max
{min
{dim
(S/Q ′),⌈dim(S/Q )
2
⌉},min
{dim(S/Q ),
⌈dim(S/Q ′)
2
⌉}}.
Proof. If one of Q , Q ′ is of dimension zero then depth(S/(Q ∩ Q ′)) = 0 and so by [5, Corol-lary 1.6] (or [7, Theorem 1.4]) sdepth(S/(Q ∩ Q ′)) = 0, that is the inequality holds trivially. Thuswe may suppose after renumbering of variables that Q is generated in variables {x1, . . . , xt} andQ ′ is generated in variables {xr+1, . . . , xp} for some integers t, r, p with 1 � r � t < p � n, or0 � r < t � n. By hypothesis we have p = n. Let D be the Stanley decomposition of S/(Q ∩ Q ′)such that sdepth(D) = sdepth(S/(Q ∩ Q ′)). Let Z1 be defined as in Lemma 2.1, that is K [Z1] is theStanley space corresponding to 1. If Z1 ⊂ {x1, . . . , xr} then by Lemma 2.1,
sdepth(D) � min
{r,
⌈n − t
2
⌉}= min
{dim
(S/Q ′),⌈dim(S/Q )
2
⌉}.
If Z1 ⊂ {xt+1, . . . , xn} we get analogously
sdepth(D) � min
{n − t,
⌈r
2
⌉}= min
{dim(S/Q ),
⌈dim(S/Q ′)
2
⌉},
which shows our inequality. �Theorem 2.3. Let Q and Q ′ be two non-zero monomial primary ideals of S with different associated primeideals. Then
sdepthS S/(
Q ∩ Q ′) � max
{min
{dim
(S/Q ′),⌈dim(S/Q ) + dim(S/(Q + Q ′))
2
⌉},
min
{dim(S/Q ),
⌈dim(S/Q ′ + dim(S/(Q + Q ′)))
2
⌉}}.
Proof. As in the proof of Proposition 2.2 we may suppose that Q is generated in variables {x1, . . . , xt}and Q ′ is generated in variables {xr+1, . . . , xp} for some integers 1 � r � t < p � n, or 0 � r < t � nbut now we have not in general p = n. Set S ′ = K [x1, . . . , xp], q = Q ∩ S ′ , q′ = Q ′ ∩ S ′ . Using Proposi-tion 2.2 we get
sdepthS
(S/
(q ∩ q′)) � max
{min
{dim
(S/q′),⌈dim(S/q)
2
⌉},min
{dim(S/q),
⌈dim(S/q′)
2
⌉}}.
By [9, Lemma 3.6] we have
sdepthS
(S/
(Q ∩ Q ′)) = sdepthS
(S/
(q ∩ q′)S
) = n − p + sdepthS ′(
S ′/(q ∩ q′)).
As in the proof of Theorem 1.5, it follows that
D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959 2949
sdepthS
(S/
(Q ∩ Q ′)) � n − p + max
{min
{r,
⌈p − t
2
⌉},min
{p − t,
⌈r
2
⌉}}
= max
{min
{n − p + r,n − t −
⌊p − t
2
⌋},min
{n − t,n − p + r −
⌊r
2
⌋}},
which is enough. �Corollary 2.4. Let Q and Q ′ be two non-zero monomial irreducible ideals of S with different associated primeideals. Then
sdepthS S/(Q ∩ Q ′) = max
{min
{dim
(S/Q ′),⌈dim(S/Q ) + dim(S/(Q + Q ′))
2
⌉},
min
{dim(S/Q ),
⌈dim(S/Q ′) + dim(S/(Q + Q ′))
2
⌉}}.
For the proof apply Theorem 1.5 and Theorem 2.3.
Corollary 2.5. Let P and P ′ be two different non-zero monomial prime ideals of S, which are not included onein the other. Then
sdepthS S/(
P ∩ P ′) = max
{min
{dim
(S/P ′),⌈dim(S/P ) + dim(S/(P + P ′))
2
⌉},
min
{dim(S/P ),
⌈dim(S/P ′) + dim(S/(P + P ′))
2
⌉}}.
Proof. For the proof apply Corollary 2.4. �Corollary 2.6. Let � be a simplicial complex in n vertices with only two different facets F , F ′ . Then
sdepth K [�] = max
{min
{∣∣F ′∣∣,⌈ |F | + |F ∩ F ′|2
⌉},min
{|F |,
⌈ |F ′| + |F ∩ F ′|2
⌉}}.
3. An illustration
Let S = K [x1, . . . , x6], Q = (x21, x2
2, x23, x2
4, x1x2x4, x1x3x4), Q ′ = (x24, x5, x6). By our Theorem 2.3 we
get
sdepth S/(
Q ∩ Q ′) � max
{min
{3,
⌈2
2
⌉},min
{2,
⌈3
2
⌉}}= max{1,2} = 2.
On the other hand, we claim that I = ((Q : w) ∩ K [x1, x2, x3]) = (x21, x2
2, x23, x1x2, x1x3) for w = x4 and
sdepth I = 1 < 2 = sdepth(Q ∩ K [x1, x2, x3]). Thus our Proposition 1.4 gives
sdepth S/(
Q ∩ Q ′) � max
{min
{3,
⌈2
2
⌉},min
{2,
⌈3
2
⌉,1
}}= 1.
In this section, we will show that sdepth(S/(Q ∩ Q ′)) = 1.First we prove our claim. Suppose that there exists a Stanley decomposition D of I with
sdepth D � 2. Among the Stanley spaces of D we have five important x21 K [Z1], x2
2 K [Z2], x23 K [Z3],
2950 D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959
x1x2 K [Z4], x1x3 K [Z5] for some subsets Zi ⊂ {x1, x2, x3} with |Zi | � 2. If Z4 = {x1, x2, x3} and Z5 con-tains x2 then the last two Stanley spaces will have a non-zero intersection and if Z1 contains x2 thenthe first and the fourth Stanley space will have non-zero intersection. Now if x2 /∈ Z5 and x2 /∈ Z1 thenthe first and the last space will intersect. Suppose that Z4 = {x1, x2}. Then x2 /∈ Z1 (resp. x1 /∈ Z2) be-cause otherwise the intersection of x1x2 K [Z4] with the first Stanley space (resp. the second one) willbe again non-zero. As |Z1|, |Z2| � 2 we get Z1 = {x1, x3}, Z2 = {x2, x3}. But x1 /∈ Z3 because otherwisethe first and the third Stanley space will contain x2
1x23, which is impossible. Similarly, x2 /∈ Z3, which
contradicts |Z3| � 2. The case Z5 = {x1, x3} gives a similar contradiction.Now suppose that Z4 = {x1, x3}. If Z5 ⊃ {x1, x2} we see that the intersection of the last two Stanley
spaces from the above five, contains x21x2x3 and if Z5 = {x2, x3} we see that the intersection of the
same Stanley spaces contains x1x2x3. Contradiction (we saw that Z5 �= {x1, x3})! Hence sdepth D � 1and so sdepth I = 1 using [5].
Next we show that sdepth S/(Q ∩ Q ′) = 1. Suppose that D′ is a Stanley decomposition ofS/(Q ∩ Q ′) such that sdepth S/(Q ∩ Q ′) = 2. We claim that D′ has the form
S/(
Q ∩ Q ′) =(⊕
v K [x5, x6])
⊕(
s⊕i=1
ui K [Zi])
for some monomials v ∈ (K [x1, . . . , x4] \ Q ), ui ∈ (Q ∩ K [x1, . . . , x4]) and Zi ⊂ {x1, x2, x3}. Indeed, letv ∈ (K [x1, . . . , x4] \ Q ). Then vx5, vx6 belong to some Stanley spaces of D′ , let us say uK [Z ], u′K [Z ′].The presence of x5 in u or Z implies that Z does not contain any xi , 1 � i � 3, otherwise uK [Z ] willbe not free over K [Z ]. Thus Z ⊂ {x5, x6}. As |Z | � 2 we get Z = {x5, x6} and similarly Z ′ = {x5, x6}.Thus vx5x6 ∈ (uK [Z ] ∩ u′K [Z ′]) and it follows that u = u′ , Z = Z ′ because the sum in D′ is direct. Itfollows that u|vx5, u|vx6 and so u|v , that is v = u f , f being a monomial in x5, x6. As v ∈ K [x1, . . . , x4]we get f = 1 and so u = v.
A monomial w ∈ (Q \ Q ′) is not a multiple of x5, x6, because otherwise w ∈ Q ′ . Suppose wbelongs to a Stanley space uK [Z ] of D′ . If u ∈ (K [x1, . . . , x4] \ Q ) then as above D′ has also aStanley space uK [x5, x6] and both spaces contains u. This is false since the sum is direct. Thusu ∈ (Q ∩ K [x1, . . . , x4]), which shows our claim.
Hence D′ induces two Stanley decompositions S/Q = ⊕v∈(K [x1,...,x4]\Q ) v K [x5, x6], Q /(Q ∩ Q ′) =⊕s
i=1 ui K [Zi], where ui ∈ (Q ∩ K [x1, . . . , x4]) and Zi ⊂ {x1, x2, x3}. Then we get the following Stanleydecompositions
Q ∩ K [x1, . . . , x3] =s⊕
i=1, ui /∈(x4)
ui K [Zi], I =s⊕
i=1, x4|ui
(ui/x4)K [Zi].
As 2 � mini |Zi | we get sdepth I � 2. Contradiction!
4. A lower bound for Stanley’s depth of some ideals
Let Q , Q ′ be two non-zero irreducible monomial ideals of S such that√
Q = (x1, . . . , xt),√
Q ′ =(xr+1, . . . , xp) for some integers r, t, p with 1 � r � t < p � n, or 0 = r < t < p � n, or 1 � r � t =p � n.
Lemma 4.1. Suppose that p = n, t = r. Then
sdepth(
Q ∩ Q ′) �⌈
r
2
⌉+
⌈n − r
2
⌉� n/2.
Proof. It follows 1 � r < p. Let f ∈ Q ∩ K [x1, . . . , xr], g ∈ Q ′ ∩ K [xr+1, . . . , xn] and M(T ) be the mono-mials from an ideal T . The correspondence ( f , g) → f g defines a map ϕ : M(Q ∩ K [x1, . . . , xr]) ×
D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959 2951
M(Q ′ ∩ K [xr+1, . . . , xn]) → M(Q ∩ Q ′), which is injective. If w is a monomial of Q ∩ Q ′ , let ussay w = f g for some monomials f ∈ K [x1, . . . , xr], g ∈ K [xr+1, . . . , xn] then f g ∈ Q and so f ∈ Qbecause the variables xi , i > r are regular on S/Q . Similarly, g ∈ Q ′ and so w = ϕ(( f , g)), that is ϕis surjective. Let D be a Stanley decomposition of Q ∩ K [x1, . . . , xr],
D: Q ∩ K [x1, . . . , xr] =s⊕
i=1
ui K [Zi]
with sdepth D = sdepth(Q ∩ K [x1, . . . , xr]) and D′ a Stanley decomposition of Q ′ ∩ K [xr+1, . . . , xn],
D′: Q ′ ∩ K [xr+1, . . . , xn] =e⊕
j=1
v j K [T j]
with sdepth D′ = sdepth(Q ′ ∩ K [xr+1, . . . , xn]). They induce a Stanley decomposition
D′′: Q ∩ Q ′ =e⊕
j=1
s⊕i=1
ui v j K [Zi ∪ T j]
because of the bijection ϕ . Thus
sdepth(
Q ∩ Q ′) � sdepth D′′ = mini, j
(|Zi| + |T j|)� min
i|Zi| + min
j|T j|
= sdepth D + sdepth D′
= sdepth(
Q ∩ K [x1, . . . , xr]) + sdepth
(Q ′ ∩ K [xr+1, . . . , xn])
=(
r −⌊
r
2
⌋)+
(n − r −
⌊n − r
2
⌋)=
⌈r
2
⌉+
⌈n − r
2
⌉� n/2. �
Remark 4.2. Suppose that n = 8, r = 1. Then by the above lemma we get sdepth(Q ∩ Q ′) �� 1
2 + � 72 = 5. Since |G(Q ∩ Q ′)| = 7 we get by [10,11] the same lower bound sdepth(Q ∩ Q ′) �
8 − � 72 � = 5. If n = 8, r = 2 then by [10,11] we have sdepth(Q ∩ Q ′) � 8 − � 12
2 � = 2 but our previouslemma gives sdepth(Q ∩ Q ′) � � 2
2 + � 62 = 4.
Lemma 4.3. Suppose that p = n. Then
sdepth(
Q ∩ Q ′) �⌈
r
2
⌉+
⌈n − t
2
⌉.
Proof. We show that
Q ∩ Q ′ = (Q ∩ Q ′ ∩ K [xr+1, . . . , xt]
)S
⊕(⊕
w
w(((
Q ∩ Q ′) : w) ∩ K [x1, . . . , xr, xt+1, . . . , xn]
)),
where w runs in the monomials of K [xr+1, . . . , xt] \ (Q ∩ Q ′). Indeed, a monomial h of S has theform h = f g for some monomials f ∈ K [xr+1, . . . , xt], g ∈ K [x1, . . . , xr, xt+1, . . . , xn]. Since Q , Q ′ are
2952 D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959
irreducible we see that h ∈ Q ∩ Q ′ either when f is a multiple of a minimal generator of Q ∩ Q ′ ∩K [xr+1, . . . , xt], or f /∈ (Q ∩ Q ′ ∩ K [xr+1, . . . , xt]) and then
h ∈ f(((
Q ∩ Q ′) : f) ∩ K [x1, . . . , xr, xt+1, . . . , xn]).
Let D be a Stanley decomposition of (Q ∩ Q ′ ∩ K [xr+1, . . . , xt])S ,
D: (Q ∩ Q ′ ∩ K [xr+1, . . . , xt]
)S =
s⊕i=1
ui K [Zi]
with sdepth D = sdepth(Q ∩ Q ′ ∩ K [xr+1, . . . , xt])S and for all w ∈ (K [xr+1, . . . , xt] \ (Q ∩ Q ′)), letD w be a Stanley decomposition of ((Q ∩ Q ′) : w) ∩ K [x1, . . . , xr, xt+1, . . . , xn],
D w : ((Q ∩ Q ′) : w
) ∩ K [x1, . . . , xr, xt+1, . . . , xn] =⊕
w
⊕j
v w j K [T w j]
with sdepth D w = sdepth(((Q ∩ Q ′) : w) ∩ K [x1, . . . , xr, xt+1, . . . , xn]). Since K [xr+1, . . . , xt] \ (Q ∩ Q ′)contains just a finite set of monomials we get a Stanley decomposition of Q ∩ Q ′ ,
D′: Q ∩ Q ′ =(
s⊕i=1
ui K [Zi])
⊕(⊕
w
⊕j
w v w j K [T w j])
,
where w runs in the monomials of K [xr+1, . . . , xt] \ (Q ∩ Q ′). Then
sdepth D′ = minw
{sdepth D, sdepth D w}= min
w
{sdepth
(Q ∩ Q ′ ∩ K [xr+1, . . . , xt]
)S,
sdepth(((
Q ∩ Q ′) : w) ∩ K [x1, . . . , xr, xt+1, . . . , xn]
)}.
But ((Q ∩ Q ′) : w) ∩ K [x1, . . . , xr, xt+1, . . . , xn] is still an intersection of two irreducible ideals and
sdepth(((
Q ∩ Q ′) : w) ∩ K [x1, . . . , xr, xt+1, . . . , xn]) �
⌈r
2
⌉+
⌈n − t
2
⌉
by Lemma 4.1. We have sdepth(Q ∩ Q ′ ∩ K [xr+1, . . . , xt]) � 1 and so
sdepth(
Q ∩ Q ′ ∩ K [xr+1, . . . , xt])
S � 1 + n − t + r
by [9, Lemma 3.6]. Thus
sdepth(
Q ∩ Q ′) � sdepth D′ �⌈
r
2
⌉+
⌈n − t
2
⌉.
Note that the proof goes even when 0 � r < t � n (anyway sdepth Q ∩ Q ′ � 1 if n = t , r = 0). �Lemma 4.4.
sdepth(
Q ∩ Q ′) � n − p +⌈
r
2
⌉+
⌈p − t
2
⌉.
D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959 2953
Proof. As usual we see that there are now (n − p) free variables and it is enough to apply [9, Lemma3.6] and Lemma 4.3. �Theorem 4.5. Let Q and Q ′ be two non-zero irreducible monomial ideals of S. Then
sdepthS
(Q ∩ Q ′) � dim
(S/
(Q + Q ′)) +
⌈dim(S/Q ′) − dim(S/(Q + Q ′))
2
⌉
+⌈
dim(S/Q ) − dim(S/(Q + Q ′))2
⌉
�⌈
dim(S/Q ′) + dim(S/Q )
2
⌉.
Proof. After renumbering of variables, we may suppose as above that√
Q = (x1, . . . , xt),√
Q ′ =(xr+1, . . . , xp) for some integers r, t, p with 1 � r � t < p � n, or 0 = r < t < p � n, or 1 � r �t = p � n. If n = p, r = 0 then
√Q ⊂ √
Q ′ and the inequality is trivial. It is enough to ap-ply Lemma 4.4 because n − p = dim(S/(Q + Q ′)), r = dim(S/Q ′) − dim(S/(Q + Q ′)), p − t =dim(S/Q ) − dim(S/(Q + Q ′)). �Remark 4.6. If Q , Q ′ are non-zero irreducible monomial ideals of S with
√Q = √
Q ′ then we havesdepthS (Q ∩ Q ′) � 1 + dim S/Q .
Example 4.7. Let S = K [x1, x2], Q = (x1), Q ′ = (x21, x2). We have
sdepth(
Q ∩ Q ′) �⌈
dim(S/Q ′) + dim(S/Q )
2
⌉=
⌈1 + 0
2
⌉= 1
by the above theorem. As Q ∩ Q ′ is not a principle ideal its Stanley depth is < 2. Thus
sdepth(
Q ∩ Q ′) = 1.
Example 4.8. Let S = K [x1, x2, x3, x4, x5], Q = (x1, x2, x23), Q ′ = (x3, x4, x5). As dim(S/(Q + Q ′)) = 0,
dim S/Q = 2 and dim S/Q ′ = 2 we get
sdepth(
Q ∩ Q ′) �⌈
dim(S/Q ′) + dim(S/Q )
2
⌉=
⌈2 + 2
2
⌉= 2
by the above theorem. Note also that
sdepth(
Q ∩ Q ′ ∩ K [x1, x2, x4, x5]) = sdepth(x1x4, x1x5, x2x4, x2x5)K [x1, x2, x4, x5] = 3,
and
sdepth(((
Q ∩ Q ′) : x3) ∩ K [x1, x2, x4, x5]
) = sdepth((x1, x2)K [x1, x2, x4, x5]
)= 4 −
⌊2
2
⌋= 3,
by [15]. But sdepth(Q ∩ Q ′) � 3 because of the following Stanley decomposition
2954 D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959
Q ∩ Q ′ = x1x4 K [x1, x4, x5] ⊕ x1x5 K [x1, x2, x5] ⊕ x2x4 K [x1, x2, x4] ⊕ x2x5 K [x2, x4, x5]⊕ x2
3 K [x3, x4, x5] ⊕ x2x3 K [x2, x3, x4] ⊕ x1x3 K [x1, x2, x3] ⊕ x1x3x4 K [x1, x2, x4, x5]⊕ x1x3x5 K [x1, x3, x5] ⊕ x2x3x5 K [x2, x3, x4, x5] ⊕ x1x2x4x5 K [x1, x2, x4, x5]⊕ x1x2
3x4 K [x1, x3, x4, x5] ⊕ x1x2x3x5 K [x1, x2, x3, x5] ⊕ x1x2x23x4 K [x1, x2, x3, x4, x5].
5. Applications
Let I ⊂ S be a non-zero monomial ideal. A. Rauf presented in [14] the following:
Question 5.1. Does it hold the inequality
sdepth I � 1 + sdepth S/I?
The importance of this question is given by the following:
Proposition 5.2. Suppose that Stanley’s Conjecture holds for cyclic S-modules and the above question has apositive answer for all monomial ideals of S. Then Stanley’s Conjecture holds for all monomial ideals of S.
For the proof note that sdepth I � 1 + sdepth S/I � 1 + depth S/I = depth I .
Remark 5.3. In [12] it is proved that Stanley’s Conjecture holds for all multigraded cycle modules overS = K [x1, . . . , x5]. If the above question has a positive answer then Stanley’s Conjecture holds for allmonomial ideals of S . Actually this is true for all square free monomial ideals of S as [13] shows.
We show that the above question holds for the intersection of two non-zero irreducible monomialideals.
Proposition 5.4. Question 5.1 has a positive answer for intersections of two non-zero irreducible monomialideals.
Proof. First suppose that Q , Q ′ have different associated prime ideals. After renumbering of variableswe may suppose as above that
√Q = (x1, . . . , xt),
√Q ′ = (xr+1, . . . , xp) for some integers r, t, p with
1 � r � t < p � n, or 0 = r < t < p � n, or 1 � r � t = p � n. Then
sdepth(
Q ∩ Q ′) � n − p +⌈
r
2
⌉+
⌈p − t
2
⌉
by Lemma 4.4. Note that
sdepth(
S/(
Q ∩ Q ′)) = n − p + max
{min
{r,
⌈p − t
2
⌉},min
{p − t,
⌈r
2
⌉}}
by Corollary 2.4. Thus
1 + sdepth(
S/(
Q ∩ Q ′)) � n − p +⌈
r
2
⌉+
⌈p − t
2
⌉� sdepth
(Q ∩ Q ′).
Finally, if Q , Q ′ have the same associated prime ideal then sdepth(Q ∩ Q ′) � 1 + dim S/Q by Re-mark 4.6 and so sdepth(Q ∩ Q ′) � 1 + sdepth S/(Q ∩ Q ′). �
D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959 2955
Next we will show that Stanley’s Conjecture holds for intersections of two primary monomialideals. We start with a simple lemma.
Lemma 5.5. Let Q , Q ′ be two primary ideals in S = K [x1, . . . , xn]. Suppose√
Q = (x1, . . . , xt) and√
Q =(xr+1, . . . , xp) for integers 0 � r � t � p � n. Then sdepth(S/(Q ∩ Q ′)) � depth(S/(Q ∩ Q ′)), that is Stan-ley’s Conjecture holds for S/(Q ∩ Q ′).
Proof. If either r = 0, or t = p then depth S/(Q ∩ Q ′) � n− p � sdepth(S/(Q ∩ Q ′)) by [9, Lemma 3.6].Now suppose that r > 0, t < p and let S ′ = K [x1, . . . , xp] and q = Q ∩ S ′ , q′ = Q ′ ∩ S ′ . Consider thefollowing exact sequence of S ′-modules
0 → S ′/(q ∩ q′) → S ′/q ⊕ S ′/q′ → S ′/
(q + q′) → 0.
By Lemma 1.2
depth(
S ′/q ⊕ S ′/q′) = min{
depth(
S ′/q),depth
(S ′/q′)}
= min{
dim(
S ′/q),dim
(S ′/q′)}
= min{r, p − t} � 1 > 0
= depth(
S ′/(q + q′)).
Thus by Depth Lemma (see e.g. [4])
depth(
S ′/q ∩ q′) = depth(
S ′/(q + q′)) + 1 = 1.
But sdepth(S ′/(q ∩ q′)) � 1 by [5, Corollary 1.6] and so
sdepth(
S/(
Q ∩ Q ′)) = sdepth(
S ′/(q ∩ q′)) + n − p � 1 + n − p
= n − p + depth(
S ′/(q ∩ q′))
= depth(
S/(
Q ∩ Q ′))by [9, Lemma 3.6]. �Theorem 5.6. Let Q , Q ′ be two non-zero irreducible ideals of S. Then sdepth(Q ∩ Q ′) � depth(Q ∩ Q ′),that is Stanley’s Conjecture holds for Q ∩ Q ′ .
Proof. By Proposition 5.4, Question 5.1 has a positive answer, so by the proof of Proposition 5.2 it isenough to know that Stanley’s Conjecture holds for S/(Q ∩ Q ′). This is given by the above lemma. �
Next we consider the cycle module given by an irredundant intersection of 3 irreducible ideals.
Lemma 5.7. Let Q 1 , Q 2 , Q 3 be three non-zero irreducible monomial ideals of S = K [x1, . . . , xn]. Then
sdepth((Q 2 ∩ Q 3)/(Q 1 ∩ Q 2 ∩ Q 3)
)� dim
(S/(Q 1 + Q 2 + Q 3)
) +⌈
dim(S/(Q 1 + Q 2)) − dim(S/(Q 1 + Q 2 + Q 3))
2
⌉
2956 D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959
+⌈
dim(S/(Q 1 + Q 3)) − dim(S/(Q 1 + Q 2 + Q 3))
2
⌉
�⌈
dim(S/(Q 1 + Q 2)) + dim(S/(Q 1 + Q 3))
2
⌉.
If Q 3 ⊂ Q 1 + Q 2 then
sdepth((Q 2 ∩ Q 3)/(Q 1 ∩ Q 2 ∩ Q 3)
)�
⌈dim(S/Q 1) + dim(S/(Q 1 + Q 3))
2
⌉.
Proof. Renumbering the variables we may assume that√
Q 1 = (x1, . . . , xt) and√
Q 2 + Q 3 =(xr+1, . . . , xp), where 0 � r � t < p � n. If t = p then
√Q 1 + Q 2 = √
Q 1 + Q 3 and the inequalityis trivial by [9, Lemma 3.6]. Let S ′ = K [x1, . . . , xp] and q1 = Q 1 ∩ S ′ , q2 = Q 2 ∩ S ′ , q3 = Q 3 ∩ S ′. Wehave a canonical injective map (q2 ∩ q3)/(q1 ∩ q2 ∩ q3) → S ′/q1. Now by Lemma 1.1, we have
S ′/q1 =⊕
uK [xt+1, . . . , xp]
and so
(q2 ∩ q3)/(q1 ∩ q2 ∩ q3) =⊕(
(q2 ∩ q3) ∩ uK [xt+1, . . . , xp]),where u runs in the monomials of K [x1, . . . , xt] \ (q1 ∩ K [x1, . . . , xt]). If u ∈ K [x1, . . . , xr] then
(q2 ∩ q3) ∩ uK [xt+1, . . . , xp] = u(q2 ∩ q3 ∩ K [xt+1, . . . , xp])
and if u /∈ K [x1, . . . , xr] then
(q2 ∩ q3) ∩ uK [xt+1, . . . , xp] = u((
(q2 ∩ q3) : u) ∩ K [xt+1, . . . , xp]).
Since (q2 ∩ q3) : u is still an intersection of irreducible monomial ideals we get by Lemma 4.3 that
sdepth((
(q2 ∩ q3) : u) ∩ K [xt+1, . . . , xp])
�⌈
dim K [xt+1, . . . , xp]/q2 ∩ K [xt+1, . . . , xp]2
⌉+
⌈dim K [xt+1, . . . , xp]/q3 ∩ K [xt+1, . . . , xp]
2
⌉.
Also we have
q2/(q1 ∩ q2) =⊕
u(q2 ∩ K [xt+1, . . . , xp]),
and it follows
S ′/(q1 + q2) ∼= (S ′/q1
)/(q2/(q1 ∩ q2)
) =⊕
u(
K [xt+1, . . . , xp]/q2 ∩ K [xt+1, . . . , xp]).Thus dim S ′/(q1 + q2) = dim K [xt+1, . . . , xp]/q2 ∩ K [xt+1, . . . , xp] and similarly
dim S ′/(q1 + q3) = dim K [xt+1, . . . , xp]/q3 ∩ K [xt+1, . . . , xp].Hence
D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959 2957
sdepth((q2 ∩ q3)/(q1 ∩ q2 ∩ q3)
)�
⌈dim(S ′/(q1 + q2))
2
⌉+
⌈dim(S ′/(q1 + q3))
2
⌉
=⌈
dim(S/(Q 1 + Q 2)) − dim(S/(Q 1 + Q 2 + Q 3))
2
⌉
+⌈
dim(S/(Q 1 + Q 3)) − dim(S/(Q 1 + Q 2 + Q 3))
2
⌉.
If Q 3 ⊂ Q 1 + Q 2 then (q2 ∩ q3) ∩ K [xt+1, . . . , xp] = q3 ∩ K [xt+1, . . . , xp] and so
sdepthS ′(q2 ∩ q3)/(q1 ∩ q2 ∩ q3) � sdepth((q2 ∩ q3) ∩ K [xt+1, . . . , xp])
= p − t −⌊
ht(q3 ∩ K [xt+1, . . . , xp])2
⌋
=⌈
p − t + dim K [xt+1, . . . , xp]/q3 ∩ K [xt+1, . . . , xp]2
⌉
=⌈
dim(S ′/q1) + dim(S ′/(q1 + q3))
2
⌉.
Now it is enough to apply [9, Lemma 3.6]. �Proposition 5.8. Let Q 1 , Q 2 , Q 3 be three non-zero irreducible ideals of S and R = S/Q 1 ∩ Q 2 ∩ Q 3 . Supposethat dim S/(Q 1 + Q 2 + Q 3) = 0. Then
sdepth R � max
{min
{sdepth S/(Q 2 ∩ Q 3),
⌈dim(S/(Q 1 + Q 2))
2
⌉+
⌈dim(S/(Q 1 + Q 3))
2
⌉},
min
{sdepth S/(Q 1 ∩ Q 3),
⌈dim(S/(Q 1 + Q 2))
2
⌉+
⌈dim(S/(Q 2 + Q 3))
2
⌉},
min
{sdepth S/(Q 1 ∩ Q 2),
⌈dim(S/(Q 3 + Q 2))
2
⌉+
⌈dim(S/(Q 1 + Q 3))
2
⌉}}.
For the proof apply Lemma 1.3 and Lemma 5.7.
Theorem 5.9. Let Q 1 , Q 2 , Q 3 be three non-zero irreducible ideals of S and R = S/(Q 1 ∩ Q 2 ∩ Q 3). Thensdepth R � depth R, that is Stanley’s Conjecture holds for R.
Proof. Applying [9, Lemma 3.6] we may reduce the problem to the case when
dim S/(Q 1 + Q 2 + Q 3) = 0.
If one of the Q i has dimension 0 then depth R = 0 and there exists nothing to show. Assume thatall Q i have dimension > 0. If one of the Q i has dimension 1 then depth R = 1 and by [5] (or [7]) weget sdepth R � 1 = depth R . From now on we assume that all Q i have dimension > 1.
If Q 1 + Q 2 has dimension 0 and Q 3 �⊂ Q 1 + Q 2 then from the exact sequence
0 → R → S/Q 1 ⊕ S/Q 2 ∩ Q 3 → S/(Q 1 + Q 2) ∩ (Q 1 + Q 3) → 0,
we get depth R = 1 by Depth Lemma and we may apply [5] (or [7]) to get as above sdepth R � 1 =depth R . If Q 3 ⊂ Q 1 + Q 2 then by Lemma 1.3, Theorem 5.6 and Lemma 5.7 we have
2958 D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959
sdepth R � min
{depth S/(Q 2 ∩ Q 3),
⌈dim(S/Q 1) + dim(S/(Q 1 + Q 3))
2
⌉}
� 1 + min{
dim S/(Q 2 + Q 3),dim S/(Q 1 + Q 3)}
= depth R
from the above exact sequence and a similar one. Thus we may suppose that Q 1 + Q 2, Q 2 + Q 3,Q 1 + Q 3 have dimension � 1. Then from the exact sequence
0 → S/(Q 1 + Q 2) ∩ (Q 1 + Q 3) → S/(Q 1 + Q 2) ⊕ S/(Q 1 + Q 3) → S/(Q 1 + Q 2 + Q 3) → 0
we get by Depth Lemma depth S/(Q 1 + Q 2) ∩ (Q 1 + Q 3) = 1. Renumbering Q i we may suppose thatdim(Q 2 + Q 3) � max{dim(Q 1 + Q 3),dim(Q 2 + Q 1)}. Using Proposition 5.8 we have
sdepth R � min
{sdepth S/Q 2 ∩ Q 3,
⌈dim(S/(Q 1 + Q 2))
2
⌉+
⌈dim(S/(Q 1 + Q 3))
2
⌉}.
We may suppose that sdepth R < dim S/Q i because otherwise sdepth R � dim S/Q i � depth R . Thususing Theorem 1.5 we get
sdepth R � min
{⌈dim S/Q 3 + dim S/(Q 2 + Q 3)
2
⌉,
⌈dim(S/(Q 1 + Q 2))
2
⌉+
⌈dim(S/(Q 1 + Q 3))
2
⌉}.
If Q 1 �⊂ √Q 3 then dim S/Q 3 > dim S/(Q 1 + Q 3) and we get
dim S/Q 3 + dim S/(Q 2 + Q 3) > dim(
S/(Q 1 + Q 2)) + dim
(S/(Q 1 + Q 3)
)because dim S/(Q 2 + Q 3) is maxim by our choice. It follows that
sdepth R �⌈
dim(S/(Q 1 + Q 2))
2
⌉+
⌈dim(S/(Q 1 + Q 3))
2
⌉� 2.
But from the first above exact sequence we get depth R = 2 with Depth Lemma, that is sdepth R �depth R .
If Q 1 �⊂ √Q 2 we note that dim S/Q 2 + dim S/(Q 2 + Q 3) > dim(S/(Q 1 + Q 2))+ dim(S/(Q 1 + Q 3))
and we proceed similarly as above with Q 2 instead Q 3. Note also that if Q 1 ⊂ √Q 2 and Q 1 ⊂√
Q 3 we get dim S/(Q 2 + Q 3) � dim S/(Q 2 + Q 1) = dim S/Q 2, respectively dim S/(Q 2 + Q 3) �dim S/(Q 3 + Q 1) = dim S/Q 3. Thus Q 1 ⊂ √
Q 3 = √Q 2 and it follows sdepth R � dim S/Q 2, which
is a contradiction. �References
[1] J. Apel, On a conjecture of R.P. Stanley, Part I – Monomial ideals, J. Algebraic Combin. 17 (2003) 39–56.[2] I. Anwar, D. Popescu, Stanley conjecture in small embedding dimension, J. Algebra 318 (2007) 1027–1031.[3] C. Biro, D.M. Howard, M.T. Keller, W.T. Trotter, S.J. Young, Interval partitions and Stanley depth, J. Combin. Theory Ser. A
(2009), in press, doi:10.1016/j.jcta.2009.07.008.[4] W. Bruns, J. Herzog, Cohen Macaulay Rings, revised edition, Cambridge University Press, Cambridge, 1996.[5] M. Cimpoeas, Stanley depth of monomial ideals in three variables, preprint, arXiv:math.AC/0807.2166, 2008.[6] M. Cimpoeas, Stanley depth of complete intersection monomial ideals, Bull. Math. Soc. Sci. Math. Roumanie 51 (99) (2008)
205–211.
D. Popescu, M.I. Qureshi / Journal of Algebra 323 (2010) 2943–2959 2959
[7] M. Cimpoeas, Some remarks on the Stanley depth for multigraded modules, Le Mathematiche LXIII (II) (2008) 165–171.[8] J. Herzog, A. Soleyman Jahan, S. Yassemi, Stanley decompositions and partitionable simplicial complexes, J. Algebraic Com-
bin. 27 (2008) 113–125.[9] J. Herzog, M. Vladoiu, X. Zheng, How to compute the Stanley depth of a monomial ideal, J. Algebra 322 (9) (2009) 3151–
3169.[10] M.T. Keller, S.J. Young, Stanley depth of squarefree monomial ideals, J. Algebra 322 (10) (2009) 3789–3792.[11] R. Okazaki, A lower bound of Stanley depth of monomial ideals, preprint, 2009.[12] D. Popescu, Stanley depth of multigraded modules, J. Algebra 321 (2009) 2782–2797.[13] D. Popescu, An inequality between depth and Stanley depth, Bull. Math. Soc. Sci. Math. Roumanie 52 (100) (2009) 377–382.[14] A. Rauf, Depth and Stanley depth of multigraded modules, Comm. Algebra, in press.[15] Y. Shen, Stanley depth of complete intersection monomial ideals and upper-discrete partitions, J. Algebra 321 (2009) 1285–
1292.[16] A. Soleyman Jahan, Prime filtrations of monomial ideals and polarizations, J. Algebra 312 (2007) 1011–1032.[17] R.P. Stanley, Linear Diophantine equations and local cohomology, Invent. Math. 68 (1982) 175–193.[18] R.H. Villarreal, Monomial Algebras, Marcel Dekker Inc., New York, 2001.